Table 2 Performance of the ML models for predicting hospitalization death in hospitalized patients with AECOPD.

Models	AUC	Sensitivity	Specificity	PPV	NPV	Accuracy	F1 score
LightGBM	0.962	0.887	0.928	0.841	0.95	0.916	0.864
GBM	0.951	0.874	0.915	0.816	0.944	0.903	0.844
XGboost	0.945	0.854	0.930	0.840	0.937	0.907	0.846
AdaBoost	0.909	0.812	0.881	0.746	0.916	0.860	0.778
RF	0.904	0.816	0.842	0.689	0.914	0.834	0.747
ET	0.893	0.774	0.863	0.709	0.899	0.836	0.740
LR	0.814	0.724	0.781	0.586	0.868	0.764	0.648
DT	0.742	0.636	0.847	0.641	0.844	0.784	0.639
KNN	0.696	0.649	0.638	0.435	0.809	0.642	0.521
ANN	0.627	0.448	0.788	0.476	0.768	0.686	0.461
SVM	0.521	0.063	0.980	0.577	0.709	0.704	0.113

The indexes represented the performance of ML models in the validation cohort. ML, machine learning; AECOPD, acute exacerbation of chronic obstructive pulmonary disease; AUC, area under the receiver-operating-characteristic curve; PPV, positive predictive value; NPV, negative predictive value; LightGBM, light gradient boosting machine; GBM, gradient boosting machine; XGboost, eXtreme gradient boosting; AdaBoost: adaptive boosting; RF, random forest; ET, extra tree; LR, logistic regression; DT, decision tree; KNN, K-nearest neighbor; ANN, artificial neutral network; SVM, support vector machine.

Quick links

Search